Direct numerical simulation (DNS) of turbulent combustion is a research area becoming ever more important and has been established as a powerful tool in combustion science to complement theory and experiment. The rapid advancement of supercomputing has lately enabled a series of interesting DNS studies with some practical relevance. This advent along with the severe lack of experimental data sets with a high degree of data completeness motivates the use of DNS for combustion modeling. Simultaneously, enormous progress is made in the utilization of DNS data as a valuable resource to develop and validate combustion models from DNS data. In this paper, several promising and useful analysis techniques are identified and reviewed. Their usefulness is demonstrated on the basis of selected modeling studies, which comprise examples of various commonly employed modeling frameworks, and which integrate DNS in a systematic way into the process of model development and validation. It is expected that the modeling routes outlined here can be used to address currently open modeling questions and to advance the fidelity of existing models.